[
https://issues.apache.org/jira/browse/KAFKA-13234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17405454#comment-17405454
]
Lucas Bradstreet commented on KAFKA-13234:
------------------------------------------
We do something similar in upgrade_test and downgrade_test. We should move this
functionality into kafka.py for use by all tests.
> Transaction system tests should check URPs between broker bounces
> -----------------------------------------------------------------
>
> Key: KAFKA-13234
> URL: https://issues.apache.org/jira/browse/KAFKA-13234
> Project: Kafka
> Issue Type: Bug
> Reporter: Jason Gustafson
> Assignee: Jason Gustafson
> Priority: Major
>
> We've seen a number of failing transaction system tests. After investigating
> a few of these, I found that this came down to transaction timeouts which
> were due to partition unavailability. The test rolls the brokers in the
> cluster, but it does not verify that URPs have cleared after restarting each
> broker. This means that it is possible for partitions to go under min isr or
> even offline. In the case of a hard bounce, the time to recover may be quite
> long because of the default session timeout of 18s. It would be good wait for
> URPs to clear instead.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
